US5179551A - Non-blocking multi-cast switching system - Google Patents

Non-blocking multi-cast switching system Download PDF

Info

Publication number
US5179551A
US5179551A US07/682,432 US68243291A US5179551A US 5179551 A US5179551 A US 5179551A US 68243291 A US68243291 A US 68243291A US 5179551 A US5179551 A US 5179551A
Authority
US
United States
Prior art keywords
connection
stage
cast
switch
networks
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
US07/682,432
Inventor
Jonathan S. Turner
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Washington University in St Louis WUSTL
Original Assignee
Washington University in St Louis WUSTL
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Washington University in St Louis WUSTL filed Critical Washington University in St Louis WUSTL
Priority to US07/682,432 priority Critical patent/US5179551A/en
Assigned to WASHINGTON UNIVERSITY, ONE BROOKINGS DRIVE, ST. LOUIS, MO 63130 A CORP OF MO reassignment WASHINGTON UNIVERSITY, ONE BROOKINGS DRIVE, ST. LOUIS, MO 63130 A CORP OF MO ASSIGNMENT OF ASSIGNORS INTEREST. Assignors: TURNER, JONATHAN S.
Priority to AU19128/92A priority patent/AU1912892A/en
Priority to PCT/US1992/002038 priority patent/WO1992017990A1/en
Application granted granted Critical
Publication of US5179551A publication Critical patent/US5179551A/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/15Interconnection of switching modules
    • H04L49/1515Non-blocking multistage, e.g. Clos
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/20Support for services
    • H04L49/201Multicast operation; Broadcast operation
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L49/00Packet switching elements
    • H04L49/25Routing or path finding in a switch fabric
    • H04L49/253Routing or path finding in a switch fabric using establishment or release of connections between ports
    • H04L49/254Centralised controller, i.e. arbitration or scheduling

Definitions

  • switching system manufacturers who have designed switching systems for high speed data networks, such as ATM networks, that employ a three-stage Benes network with fixed path routing comprising data links interconnecting each of the switches in adjacent stages.
  • high speed data packet switching networks include Clos networks and Cantor networks.
  • the internal data links In order for such a switching system to be non-blocking, the internal data links must be operated at a higher rate than the external links, i.e. the inputs and outputs to the switching system itself.
  • the ratio of a switching system's internal data rate over its internal data links to its external data rate is called a switching system's speed advantage. Minimizing the required speed advantage in order to accommodate expected data traffic is important because the speed advantage contributes directly to the switching system's cost.
  • Blocking is a term used to describe an unsuccessful attempt to transmit data over an internal data link, where the existing load on the internal data link when added to the additional load sought to be added would exceed its capacity.
  • the route selected for a virtual circuit must take into account the bandwidth in use by other virtual circuits, requiring that a path hunt be performed as in circuit switched networks. Also, as in circuit switched networks, blocking can occur if there is no path through the network with sufficient bandwidth to carry the connection.
  • a speed advantage of ⁇ is required to make an n port network non-blocking for multi-point or multi-cast traffic.
  • a five-stage network comprised of two three-stage Benes networks connected back-to-back, can be operated in a nearly non-blocking fashion with a speed advantage of three. This reduces the cost of non-blocking multi-point switching by a factor of 0.2 ⁇ n. For a 256 port switch, the cost reduction is by a factor of 3.2 and goes up to 6.4 for a 1024 port switch.
  • the same methodology may be extended to incorporate the use of Clos and/or Cantor networks, with the system being structured with a pair of networks connected back-to-back, the first network handling point-to-point traffic only while the second network is permitted to accommodate multi-cast connections.
  • the resulting switch fabric is a five-stage switching system (the adjoining stages collapsing into a single stage to reduce the total stages from six to five).
  • a 3:1 speed advantage will be all that is required for nearly non-blocking operation, with branching permitted only from stage three (the middle stage) on to the fifth stage.
  • the switching system is strictly non-blocking in that any new multi-cast connection can be added without overloading any internal data link.
  • the algorithm which accommodates the addition of a new multi-cast connection comprises the steps of: 1) identifying the most lightly loaded middle stage switch, labeled Z; 2) establishing connections between each desired multi-cast output and Z; and 3) connecting the input (labeled x) to Z.
  • a new multi-cast connection comprising the routing of a data packet from an input (x) to a plurality of outputs through the most lightly loaded middle stage switch (Z) may always be added.
  • a variation of this algorithm may also be used to augment an existing multi-cast connection by adding an output. If the output may not be conveniently added by adding an additional branch from a fourth stage or a third stage switch which already carries the connection, the augmented multi-cast connection may always be reestablished (much as a new connection) using the algorithm just mentioned. Therefore, at most, only one multi-cast connection need be re-routed, the one being augmented. All of the existing multi-cast connections may remain in place. Furthermore, in an ATM environment, the re-routing may be accomplished by changing the multi-cast translation tables in the switches comprising the third and fourth stages only, a process which requires less than one operation per switch. For example, rearranging a multi-cast connection in a 256 port switch requires making changes to at most ten of the multi-cast translation tables.
  • the technique for re-arranging an existing multi-point connection to augment it and add at least one additional end point includes the steps of first checking to see if a new branch may be added at any of the stages where branching is permitted. For a five-stage switching system comprised of two, three-stage Benes networks, these would be stages 3-5. Presuming that no additional branch could merely be added, and that re-arranging the existing multi-cast connection is required, the technique includes the steps of first identifying the most lightly loaded third stage switch (Z) and then identifying those fourth stage switches through which the connection passes and which are also connected to Z by links having sufficient capacity to carry the connection. Branches are then added from Z to each of those fourth stage switches.
  • Z most lightly loaded third stage switch
  • a multi-cast connection sought to be augmented may first be augmented through adding a branch. If no branch is available having sufficient bandwidth, then this single multi-cast connection is re-routed to accommodate the augmented multi-cast connection with minimal modification to multi-cast translation information in the switches comprising the switching system as other connections are not disturbed.
  • FIGS. 1a, b, and c depict a generalized version of a Benes, Clos, and Cantor switch network
  • FIGS. 2a and b are schematic diagrams of a three-stage Benes network switching system exemplifying blocking
  • FIG. 3 is a schematic diagram of a five-stage multi-point switching system in which branching is limited to the last three stages;
  • FIG. 4 is a schematic diagram of a five-stage multi-point switching system demonstrating the addition of a multi-cast connection without blocking
  • FIG. 5 is a schematic diagram of a five-stage multi-point switching system with a single multi-cast connection.
  • FIG. 6 is a schematic diagram of the switching system of FIG. 5 which has been re-routed in accordance with the invention in order to add an output at y.
  • the invention described herein is generally directed to high speed data packet switching systems which are utilized in various kinds of high speed data systems, including ATM networks, as is well known in the art. While the preferred embodiment disclosed and explained in some detail herein is directed to the use of Benes networks, and particularly three-stage Benes networks, it should be understood by one of ordinary skill in the art that the teachings included herein may be easily utilized with alternative high speed data packet switching networks.
  • a general depiction of a Benes network includes n inputs and n outputs with a plurality of n/d input switches and output switches, with each set of switches in each stage being connected to each set of switches in the preceding and successive stages, as shown representationally in FIG. 1a.
  • FIG. 1b Another type of high speed data packet switching network is shown in FIG. 1b as a Clos network having a similar architecture to that of the Benes except that the interior of the switch utilizes a plurality m instead of d inputs and outputs, with m being greater than d.
  • the overall system will be non-blocking if the Clos network speed advantage is greater than or equal to 1+2d/m. Obviously, for minimal cost, the speed advantage should approximate 1+2d/m.
  • FIG. 1c Still another example of a high speed data packet switching network which may be utilized in the present invention is shown in FIG. 1c and includes a Cantor network which essentially connects a plurality of Benes networks in parallel.
  • the Cantor network is non-blocking if the speed advantage is greater than or equal to 1+2(k-1)/m-2(k-2)/dm.
  • the following description of the preferred embodiment will describe in greater detail the present invention in terms of Benes networks.
  • the Clos and Cantor networks may be substituted interchangeably.
  • blocking readily occurs in a multi-cast switching system if connections are permitted to branch near the input side of the switching system. As shown in FIG. 2a, blocking can occur if there is no path through the network with sufficient bandwidth to carry a connection.
  • FIG. 2b exemplifies blocking if the multi-path routing algorithm permits early branching.
  • a single input branches in the third switch of the first stage with a 0.6 bandwidth to each of the four switches in the second stage and, then, to each of the four switches in the third stage.
  • the 0.6 multi-cast connection blocks the establishment of any connection from the c input having a bandwidth greater than 0.4.
  • a speed advantage of four is required.
  • a five-stage switching system comprised of two three-stage Benes networks connected back-to-back, with one of their connecting stages eliminated as redundant is shown in FIG. 3 in which branching is permitted for multi-point connections only in the last three stages. If this system is operated with a 3:1 speed advantage, it is always possible to add a new multi-cast connection, i.e. a multi-cast connection routed from a new input to a new set of outputs, without blocking.
  • the algorithm for adding a new multi-cast connection from an input x to a set of outputs [y 1 , . . . y r ] is as follows:
  • Each output y i has a path to more than one-half the fourth stage switches as well. Hence, there is a path from Z to each y i . By taking the union of these sets of paths, we can connect Z to every y i .
  • Z has been determined to be the most lightly loaded switch and a dotted line indicates the connections between x at the input, or first stage switch, and y 1 -y 4 outputs in the fifth stage switches.
  • a new multi-cast connection may be added to the five-stage network without blocking.
  • the network is not necessarily non-blocking if we are adding a new branch to an existing connection.
  • the existing multi-cast connection coupled with the augmented branch may be considered as a new multi-cast connection which may always be added to the five-stage network without blocking using the algorithm described above to perform the re-routing through Z.
  • This re-routing is shown by the dotted lines in FIG. 6.
  • a central processor is shown in FIGS. 6 as well understood by those of ordinary skill in the art, is used to access the translation tables resident in each switch of each stage of the overall network in order to change the routing of connections as they pass through the switch network.
  • the amount of re-routing operations may be minimized by noting that in typical implementations the multi-cast address translation information in a switch is independent of the input link that a packet is received on.
  • the switches in the last stage of the network are unaffected by the re-routing operation, with the exception, of course, of the switch to which the new output y is added.
  • the effectiveness of the operation is limited to one switch in each of the first three stages and the switches in the fourth stage. Moreover, it can be shown that at most half the switches in the fourth stage will be affected.
  • Z be the most lightly loaded third stage switch
  • S be the set of fourth stage switches that the connection passes through
  • T be the set of fourth stage switches that are connected to Z by links with at most 1-w units of bandwidth in use.
  • W be the set of switches in the fifth stage that are joined to the connection by branches passing through a switch in S-T. Add a path from Z to each switch in W and a branch to y (using step 1).
  • the time required to perform a re-arrangement is never more than the time required to establish a multi-cast connection of size ⁇ and is usually much less. As explained above, this re-arrangement is achieved by utilizing the central processor to rewrite multi-cast translation tables in the various switches being affected.
  • T 6 be the set of sixth stage switches that can be reached from Z.
  • S y be the set of seventh stage switches that the connection passes through but which are not reached through a switch in T 6 .
  • T 7 be the set of seventh stage switches that can be reached from Z.
  • S 8 be the set of eighth stage switches that the connection passes through but which are not reached through a switch in T 7 .
  • T 8 be the set of eighth stage switches that can be reached from Z.
  • S 9 be the set of ninth stage switches that the connection passes through but which are not reached through a switch in T 8 .
  • high speed data packet switching systems may be utilized in a non-blocking manner for the routing of both new multi-point connections and also augmenting of existing multi-point connections with a minimal speed advantage for each network, thereby resulting in minimal cost therefor.

Abstract

A multi-cast switching system comprised of a pair of high speed data networks, each of said data networks being either a Benes network, a Clos network, or a Cantor network, and configured to provide point-to-point switching only in the first network and multi-cast switching in the second network, may be non-blocking for adding a multi-cast connection and re-arrangeably non-blocking for augmenting an existing multi-cast connection using the algorithm which essentially consists of identifying the most lightly loaded middle stage switch, connecting the input to the middle stage switch, and connecting the outputs to the middle stage switch. A minimal speed advantage for each type of network is presented and thereby reduces the cost of each network in order to achieve non-blocking operation.

Description

This invention was made with Government support under Grant No. DCI 8600947 awarded by the National Science Foundation. The Government has certain rights in the invention.
BACKGROUND AND SUMMARY OF THE INVENTION
There are several switching system manufacturers who have designed switching systems for high speed data networks, such as ATM networks, that employ a three-stage Benes network with fixed path routing comprising data links interconnecting each of the switches in adjacent stages. Examples of other well known high speed data packet switching networks include Clos networks and Cantor networks. In order for such a switching system to be non-blocking, the internal data links must be operated at a higher rate than the external links, i.e. the inputs and outputs to the switching system itself. The ratio of a switching system's internal data rate over its internal data links to its external data rate is called a switching system's speed advantage. Minimizing the required speed advantage in order to accommodate expected data traffic is important because the speed advantage contributes directly to the switching system's cost. For example, when a three-stage Benes network switching system is used to carry point-to-point traffic, a speed advantage of three is sufficient to ensure that blocking never occurs. Blocking is a term used to describe an unsuccessful attempt to transmit data over an internal data link, where the existing load on the internal data link when added to the additional load sought to be added would exceed its capacity.
In a switching system with fixed path routing, the route selected for a virtual circuit must take into account the bandwidth in use by other virtual circuits, requiring that a path hunt be performed as in circuit switched networks. Also, as in circuit switched networks, blocking can occur if there is no path through the network with sufficient bandwidth to carry the connection.
In his prior work, the inventor herein has previously shown that a k-stage Benes network comprised switching system can be operated in a non-blocking fashion with a speed advantage of k for point-to-point traffic. See Melon and Turner "Non-Blocking Multi-Rate Distribution Network," Washington University Computer Science Department, WUCS-89-34, the disclosure of which is incorporated herein by reference.
As disclosed herein, the inventor has succeeded in extending his analysis from point-to-point switching systems to switching systems carrying multi-point or multi-cast connections. With prior art methodology, a speed advantage of √ is required to make an n port network non-blocking for multi-point or multi-cast traffic. However, using the methodology disclosed herein, a five-stage network, comprised of two three-stage Benes networks connected back-to-back, can be operated in a nearly non-blocking fashion with a speed advantage of three. This reduces the cost of non-blocking multi-point switching by a factor of 0.2√n. For a 256 port switch, the cost reduction is by a factor of 3.2 and goes up to 6.4 for a 1024 port switch. Also, the same methodology may be extended to incorporate the use of Clos and/or Cantor networks, with the system being structured with a pair of networks connected back-to-back, the first network handling point-to-point traffic only while the second network is permitted to accommodate multi-cast connections. For example, if two three-stage Benes networks are connected back-to-back, the resulting switch fabric is a five-stage switching system (the adjoining stages collapsing into a single stage to reduce the total stages from six to five). Thus, for this particular example, a 3:1 speed advantage will be all that is required for nearly non-blocking operation, with branching permitted only from stage three (the middle stage) on to the fifth stage. For new multi-cast connections, the switching system is strictly non-blocking in that any new multi-cast connection can be added without overloading any internal data link. The algorithm which accommodates the addition of a new multi-cast connection comprises the steps of: 1) identifying the most lightly loaded middle stage switch, labeled Z; 2) establishing connections between each desired multi-cast output and Z; and 3) connecting the input (labeled x) to Z. With this algorithm, a new multi-cast connection comprising the routing of a data packet from an input (x) to a plurality of outputs through the most lightly loaded middle stage switch (Z) may always be added.
A variation of this algorithm may also be used to augment an existing multi-cast connection by adding an output. If the output may not be conveniently added by adding an additional branch from a fourth stage or a third stage switch which already carries the connection, the augmented multi-cast connection may always be reestablished (much as a new connection) using the algorithm just mentioned. Therefore, at most, only one multi-cast connection need be re-routed, the one being augmented. All of the existing multi-cast connections may remain in place. Furthermore, in an ATM environment, the re-routing may be accomplished by changing the multi-cast translation tables in the switches comprising the third and fourth stages only, a process which requires less than one operation per switch. For example, rearranging a multi-cast connection in a 256 port switch requires making changes to at most ten of the multi-cast translation tables.
Generally, the technique for re-arranging an existing multi-point connection to augment it and add at least one additional end point includes the steps of first checking to see if a new branch may be added at any of the stages where branching is permitted. For a five-stage switching system comprised of two, three-stage Benes networks, these would be stages 3-5. Presuming that no additional branch could merely be added, and that re-arranging the existing multi-cast connection is required, the technique includes the steps of first identifying the most lightly loaded third stage switch (Z) and then identifying those fourth stage switches through which the connection passes and which are also connected to Z by links having sufficient capacity to carry the connection. Branches are then added from Z to each of those fourth stage switches. Then, those fifth stage switches having designated outputs for part of the connection which are not connected to the newly branched fourth stage switches are then identified and branches between each of those fifth stage switches and Z are established to complete the connection. With this technique, a multi-cast connection sought to be augmented may first be augmented through adding a branch. If no branch is available having sufficient bandwidth, then this single multi-cast connection is re-routed to accommodate the augmented multi-cast connection with minimal modification to multi-cast translation information in the switches comprising the switching system as other connections are not disturbed.
While the principal advantages and features of the subject invention have been described above, a more thorough understanding may be gained by referring to the drawings and description of the preferred embodiment which follow.
BRIEF DESCRIPTION OF THE DRAWINGS
FIGS. 1a, b, and c depict a generalized version of a Benes, Clos, and Cantor switch network;
FIGS. 2a and b are schematic diagrams of a three-stage Benes network switching system exemplifying blocking;
FIG. 3 is a schematic diagram of a five-stage multi-point switching system in which branching is limited to the last three stages;
FIG. 4 is a schematic diagram of a five-stage multi-point switching system demonstrating the addition of a multi-cast connection without blocking;
FIG. 5 is a schematic diagram of a five-stage multi-point switching system with a single multi-cast connection; and
FIG. 6 is a schematic diagram of the switching system of FIG. 5 which has been re-routed in accordance with the invention in order to add an output at y.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT
The invention described herein is generally directed to high speed data packet switching systems which are utilized in various kinds of high speed data systems, including ATM networks, as is well known in the art. While the preferred embodiment disclosed and explained in some detail herein is directed to the use of Benes networks, and particularly three-stage Benes networks, it should be understood by one of ordinary skill in the art that the teachings included herein may be easily utilized with alternative high speed data packet switching networks. For example, as shown in FIG. 1a, a general depiction of a Benes network includes n inputs and n outputs with a plurality of n/d input switches and output switches, with each set of switches in each stage being connected to each set of switches in the preceding and successive stages, as shown representationally in FIG. 1a. For Benes network switching systems, the overall system will be non-blocking if the Benes network speed advantage is greater than or equal to 2k-1-2(k-2)d, where k=logd n. Another type of high speed data packet switching network is shown in FIG. 1b as a Clos network having a similar architecture to that of the Benes except that the interior of the switch utilizes a plurality m instead of d inputs and outputs, with m being greater than d. For Clos network switching systems, the overall system will be non-blocking if the Clos network speed advantage is greater than or equal to 1+2d/m. Obviously, for minimal cost, the speed advantage should approximate 1+2d/m. Still another example of a high speed data packet switching network which may be utilized in the present invention is shown in FIG. 1c and includes a Cantor network which essentially connects a plurality of Benes networks in parallel. The Cantor network is non-blocking if the speed advantage is greater than or equal to 1+2(k-1)/m-2(k-2)/dm. The following description of the preferred embodiment will describe in greater detail the present invention in terms of Benes networks. However, as noted, the Clos and Cantor networks may be substituted interchangeably.
As shown in FIGS. 2a and b, blocking readily occurs in a multi-cast switching system if connections are permitted to branch near the input side of the switching system. As shown in FIG. 2a, blocking can occur if there is no path through the network with sufficient bandwidth to carry a connection. On the left, or in the stage one switches, is shown a set of point-to-point connections having bandwidths equal to 0.6, 0.5, 0.7, and 0.4 where for simplicity we are letting the capacity of the networks internal data links be 1. Note that given the connection shown, there is no path from the input marked a in stage one to the output marked b in stage three that has enough bandwidth remaining for a new connection with a bandwidth of 0.65.
Carrying this example to a multi-cast switching scheme, FIG. 2b exemplifies blocking if the multi-path routing algorithm permits early branching. A single input branches in the third switch of the first stage with a 0.6 bandwidth to each of the four switches in the second stage and, then, to each of the four switches in the third stage. With this example, the 0.6 multi-cast connection blocks the establishment of any connection from the c input having a bandwidth greater than 0.4. To prevent blocking from occurring in the example configuration, a speed advantage of four is required.
A five-stage switching system comprised of two three-stage Benes networks connected back-to-back, with one of their connecting stages eliminated as redundant is shown in FIG. 3 in which branching is permitted for multi-point connections only in the last three stages. If this system is operated with a 3:1 speed advantage, it is always possible to add a new multi-cast connection, i.e. a multi-cast connection routed from a new input to a new set of outputs, without blocking. The algorithm for adding a new multi-cast connection from an input x to a set of outputs [y1, . . . yr ] is as follows:
1. Find the most lightly loaded middle stage switch (that is the switch with the least amount of traffic passing through its outputs) and call it Z. The total bandwidth in use on Z's outputs is less than √n/3. Since Z has √n outputs and since a data link must be carrying a total bandwidth of more than two-thirds to block a new connection (which can have a bandwidth of at most one-third), more than half of Z's outputs have enough bandwidth available to accept a new connection. Hence more than half of the switches in the fourth stage can be reached from Z.
2. Each output yi has a path to more than one-half the fourth stage switches as well. Hence, there is a path from Z to each yi. By taking the union of these sets of paths, we can connect Z to every yi.
3. Finally, note that since branching is not allowed in the first two stages, there are paths from the input x to more than half of the second stage switches and since Z is the most lightly loaded middle stage switch, there is a path from Z to more than half of the second stage switches. Hence, there is a path from x to Z.
As shown in FIG. 4, Z has been determined to be the most lightly loaded switch and a dotted line indicates the connections between x at the input, or first stage switch, and y1 -y4 outputs in the fifth stage switches. Hence, using the methodology of the algorithm explained above, a new multi-cast connection may be added to the five-stage network without blocking. However, the network is not necessarily non-blocking if we are adding a new branch to an existing connection.
While most often such additions can be accommodated in a straightforward way, such as by merely adding a branch to the fifth stage switch from a fourth stage switch over an internal data link having sufficient free bandwidth to accommodate the connection, occasionally they require that the connection being augmented be rerouted. Hence, the network is only rearrangably non-blocking when we consider augmentation of existing multi-point connections, but the re-arrangement algorithm is quite simple because of two characteristics. First, only the connection being augmented need be re-routed. All other existing connections are not affected. Secondly, in an ATM environment, the re-routing is accomplished by changing the multi-cast translation tables in the third and fourth stage switches (for the example given of a five-stage network), which requires less than one operation per switch. For example, re-arranging a multi-cast connection in a 256 port switch requires making changes to at most ten of the multi-cast translation tables.
Consider the connections shown in FIG. 5 and suppose that the existing multi-cast connection is sought to be augmented by adding a branch to output y. The numbers next to the selected links indicate the amount of bandwidth on those links by other multi-cast connections (not shown). Note that there is no path from y to any of the fourth-stage switches that the existing multi-cast connection passes through with sufficient available bandwidth. Also, there is no path from y to the third stage switch that the multi-cast connection passes through. Since branching is not allowed in the first two stages, some re-arrangement is required to add y to the existing multi-cast connection. Assume that the switch marked Z is the most lightly loaded switch in the third stage. Effectively, the existing multi-cast connection coupled with the augmented branch may be considered as a new multi-cast connection which may always be added to the five-stage network without blocking using the algorithm described above to perform the re-routing through Z. This re-routing is shown by the dotted lines in FIG. 6. Incidentally, a central processor is shown in FIGS. 6 as well understood by those of ordinary skill in the art, is used to access the translation tables resident in each switch of each stage of the overall network in order to change the routing of connections as they pass through the switch network.
The amount of re-routing operations may be minimized by noting that in typical implementations the multi-cast address translation information in a switch is independent of the input link that a packet is received on. Thus, the switches in the last stage of the network are unaffected by the re-routing operation, with the exception, of course, of the switch to which the new output y is added. Also, there is no need to modify the multi-cast translation information in the switches which the multi-cast connection no longer passes through after re-routing. Hence, the effectiveness of the operation is limited to one switch in each of the first three stages and the switches in the fourth stage. Moreover, it can be shown that at most half the switches in the fourth stage will be affected.
Turning now to the particular example depicted in FIGS. 5 and 6, i.e. that of a five-stage Benes switching system comprised of two three-stage Benes networks, the algorithm for adding a new output y to an existing multi-cast connection, i.e. augmenting an existing multi-cast connection, is as follows.
1. If there is a branch of the connection in the fifth stage switch containing y, add y to the connection at that point. Else, if there is an available path from y to a fourth stage switch that the connection passes through, connect y at that point. Else, if there is an available path from y to the third stage switch that the connection passes through, connect y by that path.
2. Let Z be the most lightly loaded third stage switch, let S be the set of fourth stage switches that the connection passes through and let T be the set of fourth stage switches that are connected to Z by links with at most 1-w units of bandwidth in use.
Add branches from Z to all switches in S∩T.
Let W be the set of switches in the fifth stage that are joined to the connection by branches passing through a switch in S-T. Add a path from Z to each switch in W and a branch to y (using step 1).
The time required to perform a re-arrangement is never more than the time required to establish a multi-cast connection of size √ and is usually much less. As explained above, this re-arrangement is achieved by utilizing the central processor to rewrite multi-cast translation tables in the various switches being affected.
This same technique may be expanded to a five-stage Benes network which requires a speed advantage of n2/3 in order to be non-blocking for multi-cast connections. Similarly, a nine-stage network with a speed advantage of five and operated in a fashion similar to that described above is non-blocking. The algorithm for adding an additional end point y to a nine-stage network is as follows.
1. If there is a branch in the ninth stage containing y, add y to connection at that point. Else, if there is a path from y to an eighth stage switch that the connection passes through, connect y at that point. Else, if there is a path from y to a seventh stage switch that the connection passes through, connect y at that point. Else, if there is a path from y to a sixth stage switch that the connection passes through, connect y at that point. Else, if there is a path from y to a fifth stage switch that the connection passes through, connect y at that point.
2. Otherwise, let Z be most lightly loaded fifth stage switch.
Let S6 be the set of sixth stage switches that connection passes through.
Let T6 be the set of sixth stage switches that can be reached from Z.
Let Sy be the set of seventh stage switches that the connection passes through but which are not reached through a switch in T6.
Let T7 be the set of seventh stage switches that can be reached from Z.
Let S8 be the set of eighth stage switches that the connection passes through but which are not reached through a switch in T7.
Let T8 be the set of eighth stage switches that can be reached from Z.
Let S9 be the set of ninth stage switches that the connection passes through but which are not reached through a switch in T8.
Add branches from Z to every switch in (S6 ∩T6) ∪ (S7 ∩T7)∪(S8 ∩T8) ∪S9.
Through the use of these algorithms, high speed data packet switching systems may be utilized in a non-blocking manner for the routing of both new multi-point connections and also augmenting of existing multi-point connections with a minimal speed advantage for each network, thereby resulting in minimal cost therefor.
There are various changes and modifications which may be made to the invention as would be apparent to those skilled in the art. However, these changes or modifications are included in the teaching of the disclosure, and it is intended that the invention be limited only by the scope of the claims appended hereto.

Claims (14)

What is claimed is:
1. A multi-cast high speed data packet switching fabric, said fabric comprising a first network having means for handling point-to-point traffic and a second network connected back-to-back with said first network, said second network having means for handling multi-cast traffic, said first and second networks each being one of either a Benes network with a speed advantage of approximately 2k-1-2(k-2)/d, a Clos network with a speed advantage of approximately 1+2d/m, or a Cantor network with a speed advantage of approximately 1+2(k+1)/m-2(k-2)/dm, and said switching fabric having means for adding a new multi-cast connection through both networks without blocking.
2. The switching fabric of claim 1 wherein said switching fabric further comprises means for augmenting an existing connection without blocking by rearranging, at most, only said connection being augmented.
3. The switching fabric of claim 2 wherein said first and second networks are Benes networks, said first and second networks thereby comprising a five-stage switching fabric, and each of said Benes networks thereby has a speed advantage of approximately 3.
4. The switching fabric of claim 1 wherein said augmenting means includes means for augmenting said existing multi-cast connection by adding an additional connection y, said switching fabric having five stages, each of said stages having a plurality of switches, said augmenting means having means for implementing the following method including the steps of:
a. adding y to the connection if there is a branch of the connection in a fifth stage switch containing y, if there is no such fifth stage switch then
b. connecting y through an available path from y to a fourth stage switch through which the connection passes, if there is no such fourth stage switch then
c. connecting y through an available path from y to a third stage switch through which the connection passes, if there is no such third stage switch then
d. re-routing the augmented multi-cast connection through all five stages.
5. A multi-cast high speed data packet switching fabric, said fabric being comprised of a pair of three-stage Benes networks connected back-to-back, the first of said networks having means for handling point-to-point switching and the second of said networks having means for handling multi-cast traffic, each of said Benes networks having a speed advantage of approximately 2k-1-2 (k-2)/d, said switching fabric having means for adding a new multi-cast connection through both networks without blocking.
6. The switching fabric of claim 5 wherein said switching fabric further comprises means for augmenting an existing connection without blocking by rearranging, at most, only said connection being augmented.
7. The switching fabric of claim 6 wherein said augmenting means includes means for augmenting said existing multi-cast connection by adding an additional output y, said switching fabric having five stages, said augmenting means having means for implementing the following method including the steps of:
a. adding y to the connection if there is a branch of the connection in a fifth stage switch containing y, if there is no such fifth stage switch then
b connecting y through an available path from y to a fourth stage switch through which the connection passes, if there is no such fourth stage switch then
c. connecting y through an available path from y to a third stage switch through which the connection passes, if there is no such third stage switch then
d. re-routing the augmented multi-cast connection through all five stages.
8. The switching fabric of claim 7 wherein the step of re-routing includes identifying the most lightly loaded third stage switch (Z), identifying that set of fourth stage switches through which the connection passes (S), identifying that set of fourth stage switches connected to Z with a bandwidth available for handling the connection (T), adding branches of the connection from Z to all switches in S intersection T, identifying the set of (W) of fifth stage switches that are joined to the connection by branches passing through a switch in the set S-T, and adding a branch from Z to each switch in W and to y using the method of the preceding steps.
9. In a multi-cast high speed data packet switching fabric, said fabric being comprised of a pair of three-stage Benes networks connected back-to-back, the first of said networks having means for handling point-to-point switching and the second of said networks having means for handling multi-cast traffic, each of said Benes networks having a speed advantage of approximately 2k-1-2 (k-2)/d, said switching fabric thereby having five stages, the improvement comprising a method for augmenting an existing data packet connection to an additional output (y) in a fifth stage switch, said method comprising the steps of:
a. determining if a branch of the existing data packet connection terminates in the same fifth stage switch containing Y, if so, then
b. adding y internally in said fifth stage switch to said connection, if not then
c. connecting y through an available path from y to a fourth stage switch through which the connection passes, if no such fourth-stage switch then
d. connecting y through an available path from y to a third stage switch through which the connection passes, if no such third stage switch then
e. re-routing the augmented multi-cast connection through all five stages.
10. The switching fabric of claim 9 wherein the step of re-routing includes identifying the most lightly loaded third-stage switch (Z), identifying that set of fourth-stage switches through which the connection passes (S), identifying that set of fourth stage switches connected to Z with a bandwidth available for handling the connection (T), adding branches of the connection from Z to all switches in S intersection T, identifying the set of (W) of fifth stage switches that are joined to the connection by branches passing through a switch in the set S-T, and adding a branch from Z to each switch in W and to y using the method of the preceding steps.
11. In a multi-cast high speed data packet switching fabric, said fabric being comprised of a pair of three-stage Benes networks connected back-to-back, the first of said networks having means for handling point-to-point switching and the second of said networks having means for handling multi-cast traffic, each of said Benes networks having a speed advantage of approximately 2k-1-2 (k-2)/d, the switching fabric thereby having five stages, the improvement comprising a method for adding a new multi-cast connection through both networks without blocking, the method comprising the steps of:
a. identifying the most lightly loaded third stage switch (Z), adding a branch from the input of said new multi-cast connection to Z, and
b. adding a branch from Z to each of the outputs of said multi-cast connection.
12. The method of claim 11 wherein the step of adding a branch from Z to each of the outputs comprises the steps of:
a. identifying those fourth-stage switches having a path to Z with an available bandwidth sufficient to carry the connection, said switches being defined as a set A,
b. identifying those fourth-stage switches having a path to each output of the connection with a bandwidth great enough to carry the connection, said switches being defined as a set B,
c. adding a connection between Z to each output by adding branches to through those switches in the intersection of said sets A and B.
13. In a multi-cast high speed data packet switching fabric, said fabric being comprised of a pair of five-stage Benes networks connected back-to-back, the first of said networks having means for handling point-to-point switching and the second of said networks having means for handling multi-cast traffic, each of said Benes networks having a speed advantage of approximately five, the switching fabric thereby having nine stages, the improvement comprising a method for adding a new multi-cast connection through both networks without blocking, the method comprising the steps of:
a. identifying the most lightly loaded third stage switch (Z), adding a branch from the input of said new multi-cast connection to Z, and
b. adding a branch from Z to each of the outputs of said multi-cast connection.
14. The method of claim 13 wherein the step of adding a branch from Z to each of the outputs comprises the steps of:
a. if there is a branch in the ninth stage containing y, add y to connection at that point; else, if there is a path from y to an eighth stage switch that the connection passes through, connect y at that point; else, if there is a path from y to a seventh stage switch that the connection passes through, connect y at that point; else, if there is a path from y to a sixth stage switch that the connection passes through, connect y at that point; else, if there is a path from y to a fifth stage switch that the connection passes through, connect y at that point;
b. otherwise, let Z be most lightly loaded fifth stage switch and
let S6 be the sixth stage switches that connection passes through;
let T6 be the sixth stage that can be reached from Z;
let S7 be the seventh stage switches that the connection passes through but which are rot reached through a switch in T6 ;
let T7 be the seventh stage switches that can be reached from Z;
let S8 be the eighth stage switches that the connection passes through but which are not reached through a switch in T7 ;
let T8 be the eighth stage switches that can be reached from Z;
let S9 be the ninth stage switches that the connection passes through but which are not reached through a switch in T8 ; and
add branches from Z to every switch in (S6 ∩T6)∪(S7 ∩T7)∪(S8 ∩T8)∪S9.
US07/682,432 1991-04-08 1991-04-08 Non-blocking multi-cast switching system Expired - Fee Related US5179551A (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
US07/682,432 US5179551A (en) 1991-04-08 1991-04-08 Non-blocking multi-cast switching system
AU19128/92A AU1912892A (en) 1991-04-08 1992-03-12 Non-blocking multi-cast switching system
PCT/US1992/002038 WO1992017990A1 (en) 1991-04-08 1992-03-12 Non-blocking multi-cast switching system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
US07/682,432 US5179551A (en) 1991-04-08 1991-04-08 Non-blocking multi-cast switching system

Publications (1)

Publication Number Publication Date
US5179551A true US5179551A (en) 1993-01-12

Family

ID=24739673

Family Applications (1)

Application Number Title Priority Date Filing Date
US07/682,432 Expired - Fee Related US5179551A (en) 1991-04-08 1991-04-08 Non-blocking multi-cast switching system

Country Status (3)

Country Link
US (1) US5179551A (en)
AU (1) AU1912892A (en)
WO (1) WO1992017990A1 (en)

Cited By (67)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5305311A (en) * 1992-05-20 1994-04-19 Xerox Corporation Copy network providing multicast capabilities in a broadband ISDN fast packet switch suitable for use in a local area network
EP0606728A2 (en) * 1993-01-15 1994-07-20 AT&T Corp. Path hunt for efficient broadcast and multicast connections in multi-stage switching fabrics
WO1995033354A1 (en) * 1994-05-26 1995-12-07 Nokia Telecommunications Oy Bus extended tst architecture
US5495476A (en) * 1995-01-26 1996-02-27 International Business Machines Corporation Parallel algorithm to set up benes switch; trading bandwidth for set up time
US5497369A (en) * 1990-11-06 1996-03-05 Hewlett-Packard Company Multicast switch circuits
US5550818A (en) * 1994-09-19 1996-08-27 Bell Communications Research, Inc. System for wavelength division multiplexing/asynchronous transfer mode switching for network communication
US5608448A (en) * 1995-04-10 1997-03-04 Lockheed Martin Corporation Hybrid architecture for video on demand server
US5751934A (en) * 1995-12-08 1998-05-12 Electronics And Telecommunications Research Institute Non-blocking fault-tolerant gamma network for multi-processor system
US5768493A (en) * 1994-11-08 1998-06-16 International Businees Machines Corporation Algorithm for fault tolerant routing in benes networks
US5771234A (en) * 1995-12-06 1998-06-23 Industrial Technology Research Institute Method and system for ATM cell multiplexing under constant bit rate, variable bit rate and best-effort traffic
US5812526A (en) * 1995-12-21 1998-09-22 Industrial Technology Research Institute Traffic control mechanism in ATM communications network
US5841775A (en) * 1996-07-16 1998-11-24 Huang; Alan Scalable switching network
EP0885498A1 (en) * 1996-03-25 1998-12-23 I-Cube, Inc. Network switch with broadcast support
US5862329A (en) * 1996-04-18 1999-01-19 International Business Machines Corporation Method system and article of manufacture for multi-casting audio visual material
US5867653A (en) * 1996-04-18 1999-02-02 International Business Machines Corporation Method and apparatus for multi-cast based video conferencing
US5878025A (en) * 1994-08-02 1999-03-02 Fujitsu Limited ATM switch and method for switching path by ATM switch
US5879634A (en) * 1996-02-20 1999-03-09 Leco Corporation High pressure containment assembly
US5940389A (en) * 1997-05-12 1999-08-17 Computer And Communication Research Laboratories Enhanced partially self-routing algorithm for controller Benes networks
US5945922A (en) * 1996-09-06 1999-08-31 Lucent Technologies Inc. Widesense nonblocking switching networks
WO1999049628A1 (en) * 1998-03-23 1999-09-30 3Com Corporation Non-buffered, non-blocking multistage atm switch
US5987028A (en) * 1997-05-12 1999-11-16 Industrial Technology Research Insitute Multiple channel ATM switch
US6055233A (en) * 1996-10-14 2000-04-25 Samsung Electronics Co., Ltd. Augmented ring-banyan network and method for controlling routing therein
US6181162B1 (en) 1994-04-10 2001-01-30 Altera Corporation Programmable logic device with highly routable interconnect
US6184706B1 (en) 1996-04-05 2001-02-06 Altera Corporation Logic device architecture and method of operation
US6288739B1 (en) 1997-09-05 2001-09-11 Intelect Systems Corporation Distributed video communications system
WO2002003594A2 (en) * 2000-06-30 2002-01-10 Sycamore Networks, Inc. Method for managing circuits in a multistage cross connect
US6654342B1 (en) 2000-03-07 2003-11-25 Cisco Technology, Inc. Accumulating and distributing flow control information via update messages and piggybacked flow control information in other messages in a packet switching system
US6674721B1 (en) 2000-03-07 2004-01-06 Cisco Technology, Inc. Method and apparatus for scheduling packets being sent from a component of a packet switching system
US20040039787A1 (en) * 2002-08-24 2004-02-26 Rami Zemach Methods and apparatus for processing packets including accessing one or more resources shared among processing engines
US20040037322A1 (en) * 2002-08-24 2004-02-26 Vitaly Sukonik Methods and apparatus for processing packets including distributing packets across multiple packet processing engines and gathering the processed packets from the processing engines
US20040056757A1 (en) * 2001-09-27 2004-03-25 Venkat Konda Strictly nonblocking multicast multi-stage networks
US6728211B1 (en) 2000-03-07 2004-04-27 Cisco Technology, Inc. Method and apparatus for delaying packets being sent from a component of a packet switching system
US6735173B1 (en) 2000-03-07 2004-05-11 Cisco Technology, Inc. Method and apparatus for accumulating and distributing data items within a packet switching system
US6747972B1 (en) 2000-03-07 2004-06-08 Cisco Technology, Inc. Method and apparatus for reducing the required size of sequence numbers used in resequencing packets
US6757284B1 (en) 2000-03-07 2004-06-29 Cisco Technology, Inc. Method and apparatus for pipeline sorting of ordered streams of data items
US6788689B1 (en) 2000-03-07 2004-09-07 Cisco Technology, Inc. Route scheduling of packet streams to achieve bounded delay in a packet switching system
US6816492B1 (en) 2000-07-31 2004-11-09 Cisco Technology, Inc. Resequencing packets at output ports without errors using packet timestamps and timestamp floors
US6832261B1 (en) 2001-02-04 2004-12-14 Cisco Technology, Inc. Method and apparatus for distributed resequencing and reassembly of subdivided packets
US20050053061A1 (en) * 2003-09-06 2005-03-10 Venkat Konda Strictly nonblocking multicast multi-split linear-time multi-stage networks
US20050063410A1 (en) * 2003-09-06 2005-03-24 Venkat Konda Strictly nonblocking multicast linear-time multi-stage networks
US6885669B2 (en) 2001-09-27 2005-04-26 Teak Networks, Inc. Rearrangeably nonblocking multicast multi-stage networks
US6907041B1 (en) 2000-03-07 2005-06-14 Cisco Technology, Inc. Communications interconnection network with distributed resequencing
US20050129012A1 (en) * 2003-12-12 2005-06-16 Hau-Chun Ku Switch capable of controlling data packet transmission and related method
US6934760B1 (en) 2001-02-04 2005-08-23 Cisco Technology, Inc. Method and apparatus for resequencing of packets into an original ordering using multiple resequencing components
US6967926B1 (en) 2000-12-31 2005-11-22 Cisco Technology, Inc. Method and apparatus for using barrier phases to limit packet disorder in a packet switching system
US6990063B1 (en) 2000-03-07 2006-01-24 Cisco Technology, Inc. Distributing fault indications and maintaining and using a data structure indicating faults to route traffic in a packet switching system
US6993024B1 (en) * 2000-11-16 2006-01-31 Chiaro Networks, Ltd. System and method for router multicast control
US7012889B1 (en) 2000-11-02 2006-03-14 Cisco Technology, Inc. Method and apparatus for controlling input rates within a packet switching system
US7016305B1 (en) 2001-06-27 2006-03-21 Cisco Technology, Inc Method and apparatus for distributing information within a packet switching system
US7027397B1 (en) 2001-02-15 2006-04-11 Cisco Technology, Inc. Method and apparatus for accumulating and distributing traffic and flow control information in a packet switching system
US7051259B1 (en) 2002-10-08 2006-05-23 Cisco Technology, Inc. Methods and apparatus for communicating time and latency sensitive information
US7075940B1 (en) 2002-05-06 2006-07-11 Cisco Technology, Inc. Method and apparatus for generating and using dynamic mappings between sets of entities such as between output queues and ports in a communications system
US20060159078A1 (en) * 2003-09-06 2006-07-20 Teak Technologies, Inc. Strictly nonblocking multicast linear-time multi-stage networks
US20060165085A1 (en) * 2001-09-27 2006-07-27 Venkat Konda Rearrangeably nonblocking multicast multi-stage networks
US7092393B1 (en) 2001-02-04 2006-08-15 Cisco Technology, Inc. Method and apparatus for distributed reassembly of subdivided packets using multiple reassembly components
US7106693B1 (en) 2000-11-02 2006-09-12 Cisco Technology, Inc. Method and apparatus for pacing the flow of information sent from a device
US7263099B1 (en) 2002-08-14 2007-08-28 Juniper Networks, Inc. Multicast packet replication
US7269139B1 (en) 2001-06-27 2007-09-11 Cisco Technology, Inc. Method and apparatus for an adaptive rate control mechanism reactive to flow control messages in a packet switching system
US7313093B1 (en) 2002-11-26 2007-12-25 Cisco Technology, Inc. Methods and apparatus for selectively discarding packets during overload conditions
WO2008147926A1 (en) * 2007-05-25 2008-12-04 Venkat Konda Fully connected generalized butterfly fat tree networks
US7613200B1 (en) 2002-01-15 2009-11-03 Cisco Technology, Inc. Method and apparatus using a random indication to map items to paths and to recirculate or delay the sending of a particular item when a destination over its mapped path is unreachable
US7664897B2 (en) 2005-02-08 2010-02-16 Cisco Technology Inc. Method and apparatus for communicating over a resource interconnect
US7739426B1 (en) 2005-10-31 2010-06-15 Cisco Technology, Inc. Descriptor transfer logic
US20150146569A1 (en) * 2013-11-22 2015-05-28 Georg Rauh Two-Stage Crossbar Distributor and Method for Operation
US9548944B2 (en) 2014-02-27 2017-01-17 Infinera Corporation Systems, apparatus, and methods for non-blocking switch networks
US20170024346A1 (en) * 2015-07-23 2017-01-26 Cavium, Inc. Apparatus and method for on-chip crossbar design in a network switch using benes network
US11178072B2 (en) * 2015-06-10 2021-11-16 Mobileye Vision Technologies Ltd. Image processor and methods for processing an image

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4804956A (en) * 1986-03-31 1989-02-14 General Signal Corporation Rearrangeable digital signal space division switching system

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US4804956A (en) * 1986-03-31 1989-02-14 General Signal Corporation Rearrangeable digital signal space division switching system

Cited By (94)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5497369A (en) * 1990-11-06 1996-03-05 Hewlett-Packard Company Multicast switch circuits
US5305311A (en) * 1992-05-20 1994-04-19 Xerox Corporation Copy network providing multicast capabilities in a broadband ISDN fast packet switch suitable for use in a local area network
EP0606728A2 (en) * 1993-01-15 1994-07-20 AT&T Corp. Path hunt for efficient broadcast and multicast connections in multi-stage switching fabrics
EP0606728B1 (en) * 1993-01-15 2000-08-30 AT&T Corp. Path hunt for efficient broadcast and multicast connections in multi-stage switching fabrics
US6181162B1 (en) 1994-04-10 2001-01-30 Altera Corporation Programmable logic device with highly routable interconnect
GB2303522A (en) * 1994-05-26 1997-02-19 Nokia Telecommunications Oy Bus extended tst architecture
WO1995033354A1 (en) * 1994-05-26 1995-12-07 Nokia Telecommunications Oy Bus extended tst architecture
GB2303522B (en) * 1994-05-26 1999-01-27 Nokia Telecommunications Oy Bus extended tst architecture
US5878025A (en) * 1994-08-02 1999-03-02 Fujitsu Limited ATM switch and method for switching path by ATM switch
US5550818A (en) * 1994-09-19 1996-08-27 Bell Communications Research, Inc. System for wavelength division multiplexing/asynchronous transfer mode switching for network communication
US5768493A (en) * 1994-11-08 1998-06-16 International Businees Machines Corporation Algorithm for fault tolerant routing in benes networks
US5495476A (en) * 1995-01-26 1996-02-27 International Business Machines Corporation Parallel algorithm to set up benes switch; trading bandwidth for set up time
US5608448A (en) * 1995-04-10 1997-03-04 Lockheed Martin Corporation Hybrid architecture for video on demand server
US5771234A (en) * 1995-12-06 1998-06-23 Industrial Technology Research Institute Method and system for ATM cell multiplexing under constant bit rate, variable bit rate and best-effort traffic
US5751934A (en) * 1995-12-08 1998-05-12 Electronics And Telecommunications Research Institute Non-blocking fault-tolerant gamma network for multi-processor system
US5812526A (en) * 1995-12-21 1998-09-22 Industrial Technology Research Institute Traffic control mechanism in ATM communications network
US5879634A (en) * 1996-02-20 1999-03-09 Leco Corporation High pressure containment assembly
EP0885498A4 (en) * 1996-03-25 2006-04-19 Cube Inc I Network switch with broadcast support
EP0885498A1 (en) * 1996-03-25 1998-12-23 I-Cube, Inc. Network switch with broadcast support
US6184706B1 (en) 1996-04-05 2001-02-06 Altera Corporation Logic device architecture and method of operation
US6492834B1 (en) 1996-04-05 2002-12-10 Altera Corporation Programmable logic device with highly routable interconnect
US6414514B1 (en) 1996-04-05 2002-07-02 Altera Corporation Logic device architecture and method of operation
US6294928B1 (en) 1996-04-05 2001-09-25 Altera Corporation Programmable logic device with highly routable interconnect
US5867653A (en) * 1996-04-18 1999-02-02 International Business Machines Corporation Method and apparatus for multi-cast based video conferencing
US5862329A (en) * 1996-04-18 1999-01-19 International Business Machines Corporation Method system and article of manufacture for multi-casting audio visual material
US5841775A (en) * 1996-07-16 1998-11-24 Huang; Alan Scalable switching network
US5945922A (en) * 1996-09-06 1999-08-31 Lucent Technologies Inc. Widesense nonblocking switching networks
US6055233A (en) * 1996-10-14 2000-04-25 Samsung Electronics Co., Ltd. Augmented ring-banyan network and method for controlling routing therein
US5987028A (en) * 1997-05-12 1999-11-16 Industrial Technology Research Insitute Multiple channel ATM switch
US5940389A (en) * 1997-05-12 1999-08-17 Computer And Communication Research Laboratories Enhanced partially self-routing algorithm for controller Benes networks
US6288739B1 (en) 1997-09-05 2001-09-11 Intelect Systems Corporation Distributed video communications system
US6125112A (en) * 1998-03-23 2000-09-26 3Com Corporation Non-buffered, non-blocking multistage ATM switch
WO1999049628A1 (en) * 1998-03-23 1999-09-30 3Com Corporation Non-buffered, non-blocking multistage atm switch
US6757284B1 (en) 2000-03-07 2004-06-29 Cisco Technology, Inc. Method and apparatus for pipeline sorting of ordered streams of data items
US6747972B1 (en) 2000-03-07 2004-06-08 Cisco Technology, Inc. Method and apparatus for reducing the required size of sequence numbers used in resequencing packets
US6654342B1 (en) 2000-03-07 2003-11-25 Cisco Technology, Inc. Accumulating and distributing flow control information via update messages and piggybacked flow control information in other messages in a packet switching system
US6674721B1 (en) 2000-03-07 2004-01-06 Cisco Technology, Inc. Method and apparatus for scheduling packets being sent from a component of a packet switching system
US6990063B1 (en) 2000-03-07 2006-01-24 Cisco Technology, Inc. Distributing fault indications and maintaining and using a data structure indicating faults to route traffic in a packet switching system
US6907041B1 (en) 2000-03-07 2005-06-14 Cisco Technology, Inc. Communications interconnection network with distributed resequencing
US6788689B1 (en) 2000-03-07 2004-09-07 Cisco Technology, Inc. Route scheduling of packet streams to achieve bounded delay in a packet switching system
US6728211B1 (en) 2000-03-07 2004-04-27 Cisco Technology, Inc. Method and apparatus for delaying packets being sent from a component of a packet switching system
US6735173B1 (en) 2000-03-07 2004-05-11 Cisco Technology, Inc. Method and apparatus for accumulating and distributing data items within a packet switching system
US20020093952A1 (en) * 2000-06-30 2002-07-18 Gonda Rumi Sheryar Method for managing circuits in a multistage cross connect
WO2002003594A3 (en) * 2000-06-30 2002-06-13 Sycamore Networks Inc Method for managing circuits in a multistage cross connect
WO2002003594A2 (en) * 2000-06-30 2002-01-10 Sycamore Networks, Inc. Method for managing circuits in a multistage cross connect
US6816492B1 (en) 2000-07-31 2004-11-09 Cisco Technology, Inc. Resequencing packets at output ports without errors using packet timestamps and timestamp floors
US7012889B1 (en) 2000-11-02 2006-03-14 Cisco Technology, Inc. Method and apparatus for controlling input rates within a packet switching system
US7106693B1 (en) 2000-11-02 2006-09-12 Cisco Technology, Inc. Method and apparatus for pacing the flow of information sent from a device
US6993024B1 (en) * 2000-11-16 2006-01-31 Chiaro Networks, Ltd. System and method for router multicast control
US7009976B1 (en) 2000-12-31 2006-03-07 Cisco Technology, Inc. Method and apparatus for using barrier phases to synchronize processes and components in a packet switching system
US6967926B1 (en) 2000-12-31 2005-11-22 Cisco Technology, Inc. Method and apparatus for using barrier phases to limit packet disorder in a packet switching system
US6832261B1 (en) 2001-02-04 2004-12-14 Cisco Technology, Inc. Method and apparatus for distributed resequencing and reassembly of subdivided packets
US7092393B1 (en) 2001-02-04 2006-08-15 Cisco Technology, Inc. Method and apparatus for distributed reassembly of subdivided packets using multiple reassembly components
US6934760B1 (en) 2001-02-04 2005-08-23 Cisco Technology, Inc. Method and apparatus for resequencing of packets into an original ordering using multiple resequencing components
US7046627B1 (en) 2001-02-15 2006-05-16 Cisco Technology, Inc. Method and apparatus for accumulating and distributing traffic and flow control information in a packet switching system
US7027397B1 (en) 2001-02-15 2006-04-11 Cisco Technology, Inc. Method and apparatus for accumulating and distributing traffic and flow control information in a packet switching system
US7269139B1 (en) 2001-06-27 2007-09-11 Cisco Technology, Inc. Method and apparatus for an adaptive rate control mechanism reactive to flow control messages in a packet switching system
US7016305B1 (en) 2001-06-27 2006-03-21 Cisco Technology, Inc Method and apparatus for distributing information within a packet switching system
US7378938B2 (en) 2001-09-27 2008-05-27 Teak Technologies, Inc. Strictly nonblocking multicast multi-stage networks
US6868084B2 (en) 2001-09-27 2005-03-15 Teak Networks, Inc Strictly nonblocking multicast multi-stage networks
US20060165085A1 (en) * 2001-09-27 2006-07-27 Venkat Konda Rearrangeably nonblocking multicast multi-stage networks
US20040056757A1 (en) * 2001-09-27 2004-03-25 Venkat Konda Strictly nonblocking multicast multi-stage networks
US6885669B2 (en) 2001-09-27 2005-04-26 Teak Networks, Inc. Rearrangeably nonblocking multicast multi-stage networks
US7424011B2 (en) 2001-09-27 2008-09-09 Teak Technologies, Inc. Rearrangeably nonblocking multicast multi-stage networks
US20050117573A1 (en) * 2001-09-27 2005-06-02 Venkat Konda Rearrangeably nonblocking multicast multi-stage networks
US20050105517A1 (en) * 2001-09-27 2005-05-19 Venkat Konda Strictly nonblocking multicast multi-stage networks
US7613200B1 (en) 2002-01-15 2009-11-03 Cisco Technology, Inc. Method and apparatus using a random indication to map items to paths and to recirculate or delay the sending of a particular item when a destination over its mapped path is unreachable
US7075940B1 (en) 2002-05-06 2006-07-11 Cisco Technology, Inc. Method and apparatus for generating and using dynamic mappings between sets of entities such as between output queues and ports in a communications system
US7263099B1 (en) 2002-08-14 2007-08-28 Juniper Networks, Inc. Multicast packet replication
US7864769B1 (en) 2002-08-14 2011-01-04 Juniper Networks, Inc. Multicast packet replication
US7420972B1 (en) 2002-08-14 2008-09-02 Juniper Networks, Inc. Multicast packet replication
US7404015B2 (en) 2002-08-24 2008-07-22 Cisco Technology, Inc. Methods and apparatus for processing packets including accessing one or more resources shared among processing engines
US20040037322A1 (en) * 2002-08-24 2004-02-26 Vitaly Sukonik Methods and apparatus for processing packets including distributing packets across multiple packet processing engines and gathering the processed packets from the processing engines
US20040039787A1 (en) * 2002-08-24 2004-02-26 Rami Zemach Methods and apparatus for processing packets including accessing one or more resources shared among processing engines
US7304999B2 (en) 2002-08-24 2007-12-04 Cisco Technology Inc. Methods and apparatus for processing packets including distributing packets across multiple packet processing engines and gathering the processed packets from the processing engines
US7051259B1 (en) 2002-10-08 2006-05-23 Cisco Technology, Inc. Methods and apparatus for communicating time and latency sensitive information
US7313093B1 (en) 2002-11-26 2007-12-25 Cisco Technology, Inc. Methods and apparatus for selectively discarding packets during overload conditions
US20050053061A1 (en) * 2003-09-06 2005-03-10 Venkat Konda Strictly nonblocking multicast multi-split linear-time multi-stage networks
US20050063410A1 (en) * 2003-09-06 2005-03-24 Venkat Konda Strictly nonblocking multicast linear-time multi-stage networks
US20060159078A1 (en) * 2003-09-06 2006-07-20 Teak Technologies, Inc. Strictly nonblocking multicast linear-time multi-stage networks
US7424010B2 (en) 2003-09-06 2008-09-09 Teak Technologies, Inc. Strictly nonblocking multicast multi-split linear-time multi-stage networks
US20050128949A1 (en) * 2003-12-12 2005-06-16 Hau-Chun Ku Network system having a plurality of switches capable of improving transmission efficiency and method thereof
US20050129012A1 (en) * 2003-12-12 2005-06-16 Hau-Chun Ku Switch capable of controlling data packet transmission and related method
US7664897B2 (en) 2005-02-08 2010-02-16 Cisco Technology Inc. Method and apparatus for communicating over a resource interconnect
US7739426B1 (en) 2005-10-31 2010-06-15 Cisco Technology, Inc. Descriptor transfer logic
US20100172349A1 (en) * 2007-05-25 2010-07-08 Venkat Konda Fully Connected Generalized Butterfly Fat Tree Networks
WO2008147926A1 (en) * 2007-05-25 2008-12-04 Venkat Konda Fully connected generalized butterfly fat tree networks
US8170040B2 (en) * 2007-05-25 2012-05-01 Konda Technologies Inc. Fully connected generalized butterfly fat tree networks
US20150146569A1 (en) * 2013-11-22 2015-05-28 Georg Rauh Two-Stage Crossbar Distributor and Method for Operation
US9614787B2 (en) * 2013-11-22 2017-04-04 Siemens Aktiengesellschaft Two-stage crossbar distributor and method for operation
US9548944B2 (en) 2014-02-27 2017-01-17 Infinera Corporation Systems, apparatus, and methods for non-blocking switch networks
US11178072B2 (en) * 2015-06-10 2021-11-16 Mobileye Vision Technologies Ltd. Image processor and methods for processing an image
US20170024346A1 (en) * 2015-07-23 2017-01-26 Cavium, Inc. Apparatus and method for on-chip crossbar design in a network switch using benes network
US9916274B2 (en) * 2015-07-23 2018-03-13 Cavium, Inc. Apparatus and method for on-chip crossbar design in a network switch using benes network

Also Published As

Publication number Publication date
AU1912892A (en) 1992-11-02
WO1992017990A1 (en) 1992-10-15

Similar Documents

Publication Publication Date Title
US5179551A (en) Non-blocking multi-cast switching system
JP3163069B2 (en) How to upgrade the switch structure
US5402415A (en) Multicast virtual circuit switch using cell recycling
EP0226632B1 (en) A packet switching network with multiple packet destinations
JP2996353B2 (en) Multipath self-routing switching network for switching asynchronous time division multiplex cells.
US5471460A (en) ATM switching system with unit switch selecting function for avoiding faulty unit switch
JPH07202942A (en) Packet switchboard
US6154449A (en) Switchless network
US7245582B1 (en) Protection switching in a multi-stage switch fabric
US5864552A (en) Rearrangeable non-blocking switching network
US6147991A (en) Scalable high speed packet switch using packet diversion through dedicated channels
US4905224A (en) Sorting unit for a switching node comprising a plurality of digital switching matrix networks for fast, asynchronous packet switching networks
EP0734188B1 (en) Multistage connection switch
KR880004661A (en) Switching system
JP2679706B2 (en) Self-routing control system
US5414415A (en) Cross-connect apparatus capable of avoiding a superfluous detour route therein
JPH02206939A (en) Self-routing switch network
US6775248B1 (en) Programmable bandwidth allocation between send and receive in a duplex communication path
KR100246565B1 (en) Switching network in exchanger
JP3229841B2 (en) ATM communication network
JPH03159338A (en) Self-routing switching network and switching network adding method in the same
US6868085B1 (en) Method for parallel searching a look-up table
JPS63135039A (en) Path selection control method for waiting type switching network
SU1499367A1 (en) Routing device
JP2756604B2 (en) Self-routing switch network

Legal Events

Date Code Title Description
AS Assignment

Owner name: WASHINGTON UNIVERSITY, ONE BROOKINGS DRIVE, ST. LO

Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNOR:TURNER, JONATHAN S.;REEL/FRAME:005671/0048

Effective date: 19910408

REMI Maintenance fee reminder mailed
LAPS Lapse for failure to pay maintenance fees
FP Lapsed due to failure to pay maintenance fee

Effective date: 19970115

STCH Information on status: patent discontinuation

Free format text: PATENT EXPIRED DUE TO NONPAYMENT OF MAINTENANCE FEES UNDER 37 CFR 1.362